Разработка и исследование мер близости для прецедентной идентификации фрагментов рукописного текста

Abstract

In the article introduced and investigated proximity measures for precedential identification of words on handwritten text images. Main problem was formulated; methodology of various proximity measures comparative analysis was developed. Result of computational experiments on MatLab software for pixel comparison proximity measure, fragment projection mapping proximity measure and for subband proximity measure are presented. Subband proximity measure based on energy comparison and react on normalized Fourier transform differences in defined frequency slots. The efficiency of reviewed proximity measures is illustrated by calculated errors probability.